Purpose To develop a deep learning algorithm to automatically assess the posterior fossa on first-trimester US screening scans and identify open spina bifida (OSB) and cystic posterior fossa (CPF) anomalies. Materials and Methods This was the retrospective part of an international study involving 10 fetal medicine centers. Normal and abnormal (OSB, CPF anomaly) midsagittal fetal brain US images acquired between 11 and 14 weeks of gestation (July 2009-January 2024) with confirmed diagnosis at follow-up were evaluated. Images were manually annotated to delineate the posterior fossa. The dataset was split into a training/validation set (70%) and internal test set (30%). Three convolutional neural networks were trained via threefold cross-validation on the training/validation set, with predictions on the internal test set obtained by ensemble averaging across folds. Model performance in detecting OSB and CPF anomalies was evaluated for the whole cohort and for fetuses with OSB or CPF anomalies separately. Results Images from 251 fetuses were analyzed (mean gestational age [±SD], 12.7 weeks ± 0.65; 150 normal and 101 abnormal [43 OSB and 58 CPF anomalies] images). On the internal test, the MobileNetV3 Large Weights achieved the best performance: area under the receiver operating characteristic curve, 0.94 (95% CI: 0.88, 0.99); accuracy, 88% (67 of 76); recall, 81% (25 of 31); specificity, 93% (42 of 45); precision, 89% (25 of 28); negative predictive value, 88% (42 of 48); and F1 score, 0.85. OSB was classified more accurately (93% [52 of 56] vs 88% [57 of 65]; P = .38) and with higher recall (91% [10 of 11] vs 75% [15 of 20]), although the difference was not significant (P = .38). Conclusion MobileNetV3 Large Weights accurately assessed the fetal posterior fossa between 11 and 14 weeks of gestation, distinguishing normal images from those showing OSB or CPF anomalies. Clinical trial registration no. NCT0579047 Keywords: Artificial Intelligence, First Trimester Ultrasound Screening, Fetal Brain Anomalies, Deep Learning Supplemental material is available for this article. © RSNA, 2026 See also commentary by Rafful in this issue.

Abstract Objectives To assess the diagnostic potential of magnetic resonance imaging (MRI) radiomics and machine learning models using T2-weighted and contrast-enhanced (CE)-T1-weighted images, individually and combined, to predict the invasiveness of pituitary neuroendocrine tumors (PitNETs). Materials and methods Patients with macro-PitNETs were retrospectively enrolled from 2019 to 2022. Radiomic features were extracted from manually segmented lesions on preoperative T2-weighted and CE-T1-weighted images and, after a feature selection step, used to assess invasiveness, defined following Trouillas’ classification. Five machine learning models (logistic regression, random forest, gradient boosting, AdaBoost, XGBoost) were trained using CE-T1-weighted, T2-weighted, and CE-T1-weighted plus T2-weighted features. Performance was evaluated on a test set using the area under the receiver operating characteristic curve (AUC). Results Two hundred patients were included in the study: 95 PitNETs were noninvasive (74 grade 1a; 21 grade 1b) and 105 invasive (70 grade 2a; 35 grade 2b). A total of 102 radiomic features were extracted per sequence. The best-performing model was the XGBoost, using five combined CE-T1-weighted and T2-weighted features, with an AUC of 0.85 (95% confidence interval: 0.75‒0.95). Lower AUC values were obtained for logistic regression using CE-T1-weighted images (0.80) and AdaBoost using T2-weighted images (0.78). Conclusion The XGBoost model, incorporating tumor shape, texture, and first-order features extracted from both CE-T1-weighted and T2-weighted MRI, showed high performance in predicting PitNETs invasiveness. This radiomic model might help identify tumors with a higher risk of disease persistence, recurrence, or progression. Relevance statement The radiomic model based on contrast-enhanced T1-weighted and T2-weighted MRI demonstrated high discriminative ability in predicting invasiveness of pituitary neuroendocrine tumors and could aid in identifying tumors that may be at higher risk for recurrence or progression, ultimately improving patient outcomes through personalized treatment strategies. Key Points Pituitary neuroendocrine tumors (PitNETs) represent a significant challenge in clinical practice. Accurate preoperative prediction of PitNET invasiveness is crucial for surgery and prognosis. Contrast-enhanced T1-weighted and T2-weighted MRI-based radiomic model effectively predicts PitNET invasiveness. The developed radiomic model could help optimize individualized treatment decisions before surgery. Graphical Abstract

Background: Spread through air spaces (STAS) is a prognostic factor for survival in non-small cell lung cancer (NSCLC), but the chance to identify it before surgery remains challenging. In this study, we consider clinical, pathological and metabolic factors, with the aim to identify possible STAS predictors in NSCLC. Methods: Clinical and pathological characteristics of patients who underwent anatomical lung resection from 1 January 2018 to 31 December 2023 were retrospectively reviewed and analyzed. Patients with GGO, AIS, or MIA tumors, metastases, or undergoing neoadjuvant therapy were excluded. The parameters assessed by 18F-FDG PET/CT were: SUVmax, SUVmean, SUVpeak, TLG and MTV. The primary endpoint was the association between clinical, metabolic, and pathologic characteristics with the presence of STAS. A univariate and multivariate logistic regression model was developed to identify independent predictors of STAS. Results: The final analysis was conducted on 224 patients. Non-lepidic-acinar adenocarcinomas showed a statistically significant higher risk of STAS than the lepidic-acinar histotype: 20.8% vs. 4.3%, p = 0.003, OR 5.87, 95%CI 1.83-18.78. Furthermore, tumor grade was significantly associated with STAS: 4.6% in G1-G2 tumors vs. 23.5% in G3 tumors (p = 0.001, OR 5.79, 95%CI 2.12-15.78); as well as lymph vascular invasion (p = 0.034) and tumor size >2 cm (p = 0.017). Multivariate analysis confirmed high tumor grade as an independent risk factor (OR 4.73, 95%CI 1.16-19.23, p = 0.030). Conclusions: In our study, risk of STAS is not correlated with metabolic parameters, while adenocarcinoma subtypes and tumors grading seem to stratify the risk of STAS occurrence. These results may be consolidated in an external validation analysis to possibly better plan the extent of lung resection.

Management of autosomal dominant polycystic kidney disease (ADPKD) might take advantage of the use of new tools to predict risk of progression towards end stage kidney disease. The aim of this study is to explore the potential of radiomic features obtained from computed tomography (CT) scans for the prediction of kidney function decline over time of ADPKD patients. We retrospectively selected a cohort of 58 ADPKD patients who routinely underwent CT scan for total kidney volume (TKV) assessment from February 2020 to March 2021. An expert radiologist generated a region-of-interest segmentation for cystic kidneys from which we extracted 217 radiomic features. In a subgroup of 51 patients with at least 3 serum creatinine measurements, on the basis of eGFR we identified 26 rapid progressors to ESKD (>3 ml/min/year), and we developed a radiomic model to discriminate rapid from not rapid progressors. Area under the curve (AUC) of the receiver operating characteristic (ROC) and sensitivity were employed to evaluate models’ performance. The most statistically significant radiomic feature (F_cm.corr) (p-value = 0.04) associated with rapid progression showed an AUC (95% CI) of 0.78 (0.65–0.90) and a sensitivity of 0.92 (0.78–0.98). On the contrary, the logistic regression model based on the ht-TKV presented a lower AUC (95% CI) of 0.65 (0.49–0.80), with a sensitivity 0.62 (0.42–0.78). We developed a model based on the radiomic feature F_cm.corr that was able to discriminate rapid progressors. Further studies on larger cohort are warranted to validate our findings and to confirm the role of radiomics in ADPKD management.

Background:Management of autosomal dominant polycystic kidney disease (ADPKD) might take advantage of the use of new tools to predict risk of progression towards end-stage kidney disease (ESKD). The aim of this study is to explore the potential of radiomic features obtained from computed tomography (CT) scans for the prediction of kidney function decline over time of ADPKD patients. Methods:We retrospectively selected a cohort of 58 ADPKD patients who routinely underwent CT scan for total kidney volume (TKV) assessment from February 2020 to March 2021. An expert radiologist generated a region-of-interest segmentation for cystic kidneys from which we extracted 217 radiomic features. In a subgroup of 51 patients with at least three serum creatinine measurements, on the basis of estimated glomerular filtration rate we identified 26 rapid progressors to ESKD (>3 mL/min/1.73 m2/year), and we developed a radiomic model to discriminate rapid from non-rapid progressors. Area under the curve (AUC) of the receiver operating characteristic (ROC) and sensitivity were employed to evaluate models' performance. Results:The most statistically significant radiomic feature (F_cm.corr) (P-value = .04) associated with rapid progression showed an AUC (95% confidence interval) of 0.78 (0.65-0.90) and a sensitivity of 0.92 (0.78-0.98). On the contrary, the logistic regression model based on the height-adjusted TKV (ht-TKV) presented a lower AUC (95% confidence interval) of 0.65 (0.49-0.80), with a sensitivity 0.62 (0.42-0.78). Conclusions:We developed a model based on the radiomic feature F_cm.corr that was able to discriminate rapid progressors. Further validation studies on larger and external cohort are warranted to corroborate our findings and to confirm the role of radiomics in ADPKD management.

There is a clinical need to identify early predictors for response to neoadjuvant chemotherapy (NAC) in patients with gastric and gastroesophageal junction cancer (GC and GEJC). Radiomics involves extracting quantitative features from medical images. This study aimed to apply radiomics to build prediction models for the response to NAC. All consecutive patients with non-metastatic GC and GEJC undergoing NAC and surgical resection in an Italian high-volume referral center between 2005 and 2021 were considered eligible. In patients selected, the CT scans performed upon staging were reviewed to segment the tumor and extract radiomic features using MODDICOM. The primary endpoint was to develop and validate radiomic-based predictive models to identify major responders (MR: tumor regression grade TRG 1–2) and non-responders (NR: TRG 4–5) to NAC. Following an initial feature selection, radiomic and combined radiomic-clinicopathologic prediction models were built for the MR or NR status based on logistic regressions. Internal validation was performed for each model. Radiomic models (in the entire case series and according to NAC regimens) were evaluated using the receiver operating characteristic area under the curve (AUC), sensitivity, and negative predictive value (NPV). The study included 77 patients undergoing NAC and subsequent tumor resection. The MR prediction model after all types of NAC (AUC of 0.876, CI 95

Breast cancer (BC) is a major global health issue with significant heterogeneity among its subtypes. Neoadjuvant treatment (NAT) has been extended to include early BC patients, particularly those with HER2 + and triple-negative subtypes, to achieve pathological complete response and improve long-term outcomes. However, disease recurrence remains a challenge, highlighting the need for predictive biomarkers. This study evaluates the role of radiomics from pre-treatment breast MRI, integrated with clinical and radiological variables, in predicting early disease recurrence (EDR) after NAT. A retrospective analysis was conducted on 238 BC patients treated with NAT and assessed using pre- and post-treatment breast MRI. Radiomic features were extracted and combined with clinical and radiological data to develop predictive models for EDR. Models were evaluated using AUC, accuracy, sensitivity, and specificity metrics. The radiological-radiomic model, which integrated pre-treatment MRI radiomics with RECIST response data, demonstrated the highest predictive performance for EDR (AUC 0.77, sensitivity 0.85). Internal validation confirmed the robustness of the model. Combining radiomic features from pre-NAT MRI with RECIST response evaluation from post-NAT MRI enhances the prediction of EDR in BC patients, supporting precision medicine in treatment strategies and follow-up planning. Further validation on larger cohorts is needed to confirm these findings.

e17561 Background: BRCA 1/2 genes mutation identification enables women to opt for effective risk-reducing surgeries. Current indications for BRCA testing based on clinical-criteria/family-history based a priori BRCA probability thresholds are ineffective as most of the carriers remain undiagnosed. We already showed the feasibility of performing a radiomic analysis of ultrasound images of normal ovaries to predict BRCA 1/2 genes status, with performances on the testing set reasonably encouraging. Moreover, we performed a cost-effective analysis showing that combining clinical criteria with the radiogenomic model would have a massive effect after only one generation in detecting carriers in the general population with only a small cost increment. The present study aims at improving the preprocessing and modelling pipeline on a larger dataset and at validating the predictive model prospectively in a multicenter study. In this abstract we will present preliminary results from the retrospective phase. Methods: We conducted a retrospective multicenter observational study aimed at collecting ultrasound images of healthy ovaries with known BRCA status. Patients referring to participating center from January until December 2023 fulfilling the following selection criteria: 1. Availability of gBRCA1/2 test results; 2. Transvaginal ultrasound performed providing at least one picture of one healthy ovary. Healthy ovaries were manually segmented on ultrasound images. Image preprocessing steps were performed for speckle noise reduction, intensities normalization and calipers’ correction. Radiomic features were extracted from the segmented ovaries and the cohort was divided into training (70%) and validation (30%) sets. Radiomics features were selected with Recursive Feature Elimination (RFE) on the training set and used for the classification of BRCA status using different Machine Learning classifiers. The performances were evaluated considering area under the receiver operating characteristics curve (AUC). Results: 481 patients (282 BRCA-mutated and 199 wild-type) were analysed. 12 statistical and textural radiomics features were selected and used for classification. The Random-Forest radiomics model shows the best performance with an AUC of 0.74 in the validation set. Conclusions: The information on BRCA carrier status may allow in the future to benefit from the reduction in the number of cancer cases and related economic savings deriving from avoiding cost of genetic testing screening-based proposed. Testing one generation with radiogenomics screening could help reducing the burden of BRCA-related cancers and provide anamnestic information for subsequent generations. Future work will integrate the radiomics predictions with age and familiarity implementing a clinical-radiomics model. Clinical trial information: NCT05769517 .

OBJECTIVE:Although artificial intelligence (AI) is increasingly being applied to ultrasound imaging in gynecology, efforts to synthesize the available evidence have been inadequate. The aim of this systematic review was to summarize and evaluate the literature on the role of AI applied to ultrasound imaging in benign gynecological disorders. METHODS:Web of Science, PubMed and Scopus databases were searched from inception until August 2024. Inclusion criteria were studies applying AI to ultrasound imaging in the diagnosis and management of benign gynecological disorders. Studies retrieved from the literature search were imported into Rayyan software and quality assessment was performed using the Quality Assessment Tool for Artificial Intelligence-Centered Diagnostic Test Accuracy Studies (QUADAS-AI). RESULTS:Of the 59 studies included, 12 were on polycystic ovary syndrome (PCOS), 11 were on infertility and assisted reproductive technology, 11 were on benign ovarian pathology (i.e. ovarian cysts, ovarian torsion, premature ovarian failure), 10 were on endometrial or myometrial pathology, nine were on pelvic floor disorder and six were on endometriosis. China was the most highly represented country (22/59 (37.3%)). According to QUADAS-AI, most studies were at high risk of bias for the subject selection domain (because the sample size, source or scanner model was not specified, data were not derived from open-source datasets and/or imaging preprocessing was not performed) and the index test domain (AI models were not validated externally), and at low risk of bias for the reference standard domain (the reference standard classified the target condition correctly) and the workflow domain (the time between the index test and the reference standard was reasonable). Most studies (40/59) developed and internally validated AI classification models for distinguishing between normal and pathological cases (i.e. presence vs absence of PCOS, pelvic endometriosis, urinary incontinence, ovarian cyst or ovarian torsion), whereas 19/59 studies aimed to automatically segment or measure ovarian follicles, ovarian volume, endometrial thickness, uterine fibroids or pelvic floor structures. CONCLUSION:The published literature on AI applied to ultrasound in benign gynecological disorders is focused mainly on creating classification models to distinguish between normal and pathological cases, and on developing models to automatically segment or measure ovarian volume or follicles. © 2025 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

BACKGROUND:Tumor response to preoperative chemotherapy is a key prognostic factor for colorectal liver metastases (CRLM), but an accurate non-invasive assessment remains an unmet need. OBJECTIVE:To assess the contribution of radiomic analysis of preoperative, post-chemotherapy, gadoxetic acid (Gd-EOB-DTPA)-enhanced MRI to the non-invasive prediction of the pathologic response to systemic therapy of CRLM. METHODS:This retrospective bi-institutional study included all consecutive patients undergoing resection for CRLM (2018-2021) after preoperative oxaliplatin/irinotecan-based chemotherapy. We selected patients with stable disease or partial response at the last restaging, and with post-chemotherapy MRI performed ≤60 days before surgery. The largest CRLM (≥10 mm) was analyzed. Pathologic response was evaluated according to the TRG. The tumor (Tumor-VOI) was manually segmented on the portal venous phase (PVP) and hepatobiliary phase (HBP) sequences; a 5-mm ring of peritumoral tissue was automatically generated (Margin-VOI) and manually corrected. The predictive models underwent internal validation. RESULTS:Overall, 162 patients (median age 62.5 years, 102 men) were evaluated. Of the 131 patients with a radiologic partial response, 59 (45 %) had no tumor regression at pathology (TRG4-5). The model including both clinical variables and radiomic features extracted from the Tumor-VOI/Margin-VOI of PVP and HBP achieved the best performances: at validation, Accuracy = 0.773, Sensitivity = 0.724, Specificity = 0.812, and ROC-AUC = 0.860. The combined clinical-radiomic model outperformed the pure clinical one (p < 0.001). The features extracted from the Tumor-VOI in PVP and Margin-VOI in HBP had the highest impact. CONCLUSION:The addition of radiomic features extracted from the PVP and HBP of post-chemotherapy Gd-EOB-DTPA-enhanced MRI enhanced standard radiologic and clinical assessment of CRLM response to chemotherapy, providing a reliable non-invasive assessment of TRG.

Purpose: To explore the correlation between radiomics features extracted from OCT angiography (OCTA) of epiretinal membranes (ERMs) and baseline best-corrected visual acuity (BCVA). Design: Retrospective observational monocentric study. Participants: Eighty-three eyes affected by idiopathic ERMs, categorized into low (<= 70 letters) and high (70 letters) BCVA groups. Methods: The central 3 x 3 mm2 crop of structural and vascular en-face OCTA scans of superficial and deep retina slab, and choriocapillaris of each eye was selected. PyRadiomics was used to extract 86 features belonging to 2 different families: intensity-based statistical features describing the gray-level distribution, and textural features capturing the spatial arrangement of pixels. By employing a greedy strategy, 4 radiomic features were selected to build the final logistic regression model. The ability of the model to discriminate between low and high baseline BCVA was quantified in terms of area under the receiver operating characteristics curve (AUC). Main Outcome Measures: The 4 selected informative radiomic features were as follows: the difference average (glcm_DifferenceAverage), quantifying the average difference in gray-level between neighboring pixels; the informational measure of correlation (glcm_Imc1), giving information about the spatial correlation of pixel intensities inside the image; the long run low gray-level emphasis (glrlm_LongRunLowGrayLevelEmphasis), highlighting long segments of low gray-level values within the image; and the large area emphasis (glszm_LargeAreaEmphasis), which quantifies the tendency for larger zones of uniform intensity to occur. Results: No features exhibited a statistically significant difference between low and high BCVA values for the superficial and deep retinal slabs. Conversely, in the choriocapillaris layer, the glcm_DifferenceAverage and glcm_Imc1 features were significantly higher in the high BCVA group (P- 0.047), whereas higher values for the glrlm_LongRunLowGrayLevelEmphasis and glszm_LargeAreaEmphasis were associated with the low BCVA group (P- 0.047). Overall, these radiomic features predicted BCVA with an AUC (95% confidence interval) of 0.74 (0.63e0.85) and sensitivity/specificity of 0.67/0.75. During the cross-validation, the metrics remained stable. Conclusions: Radiomics features of the choriocapillaris in idiopathic ERMs showed a correlation with BCVA, with lower structural complexity and higher homogeneity, together with the presence of homogeneous areas with low- intensity pixel values, reflecting flow voids due to reduced microvascular perfusion, and were correlated with lower visual acuity. Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article. Ophthalmology Science 2025;5:100716 (c) 2025 by the American Academy of Ophthalmology. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

OBJECTIVE:The primary aim was to identify radiomics ultrasound features that can distinguish between benign and malignant adnexal masses with solid ultrasound morphology, and between primary malignant (including borderline and primary invasive) and metastatic solid ovarian masses, and to develop ultrasound-based machine learning models that include radiomics features to discriminate between benign and malignant solid adnexal masses. The secondary aim was to compare the discrimination performance of our newly developed radiomics models with that of the Assessment of Different NEoplasias in the adneXa (ADNEX) model and that of subjective assessment by an experienced ultrasound examiner. METHODS:This was a retrospective, observational single-center study conducted at Fondazione Policlinico Universitario A. Gemelli IRCC, in Rome, Italy. Included were patients with a histological diagnosis of an adnexal tumor with solid morphology according to International Ovarian Tumor Analysis (IOTA) terminology at preoperative ultrasound examination performed in 2014-2020, who were managed with surgery. The patient cohort was split randomly into training and validation sets at a ratio of 70:30 and with the same proportion of benign and malignant tumors in the two subsets, with malignant tumors including borderline, primary invasive and metastatic tumors. We extracted 68 radiomics features, belonging to two different families: intensity-based statistical features and textural features. Models to predict malignancy were built based on a random forest classifier, fine-tuned using 5-fold cross-validation over the training set, and tested on the held-out validation set. The variables used in model-building were patient age and radiomics features that were statistically significantly different between benign and malignant adnexal masses and assessed as not redundant based on the Pearson correlation coefficient. We evaluated the discriminative ability of the models and compared it to that of the ADNEX model and that of subjective assessment by an experienced ultrasound examiner using the area under the receiver-operating-characteristics curve (AUC) and classification performance by calculating sensitivity and specificity. RESULTS:In total, 326 patients were included and 775 preoperative ultrasound images were analyzed. Of the 68 radiomics features extracted, 52 differed statistically significantly between benign and malignant tumors in the training set, and 18 uncorrelated features were selected for inclusion in model-building. The same 52 radiomics features differed significantly between benign, primary malignant and metastatic tumors. However, the values of the features manifested overlapped between primary malignant and metastatic tumors and did not differ significantly between them. In the validation set, 25/98 (25.5%) tumors were benign and 73/98 (74.5%) were malignant (6 borderline, 57 primary invasive, 10 metastatic). In the validation set, a model including only radiomics features had an AUC of 0.80, sensitivity of 0.78 and specificity of 0.76 at an optimal cut-off for risk of malignancy of 68%, based on Youden's index. The corresponding results for a model including age and radiomics features were AUC of 0.79, sensitivity of 0.86 and specificity of 0.56 (cut-off 60%, based on Youden's index), while those of the ADNEX model were AUC of 0.88, sensitivity of 0.99 and specificity of 0.64 (at a 20% risk-of-malignancy cut-off). Subjective assessment had a sensitivity of 0.99 and specificity of 0.72. CONCLUSIONS:Our radiomics model had moderate discriminative ability on internal validation and the addition of age to this model did not improve its performance. Even though our radiomics models had discriminative ability inferior to that of the ADNEX model, our results are sufficiently promising to justify continued development of radiomics analysis of ultrasound images of adnexal masses. © 2024 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Abstract Background During their lifetime, patients with Crohn’s disease (CD) frequently develop strictures as a complication, where various degrees of inflammation and fibrosis coexist. Distinguishing between inflammatory and fibrotic strictures is crucial for deciding on surgery or treatment changes, yet it is currently challenging in clinical practice. Cross-sectional imaging techniques (e.g. magnetic resonance imaging MRI) provide a reasonably accurate estimate of bowel wall fibrosis. However, only histopathological examination can determine its true extent. Radiomics uses artificial intelligence to analyze quantitative features from diagnostic images and can predict clinical outcomes. This study aims to assess the role of radiomics in classifying patients according to the degree (no/mild vs severe) of intestinal fibrosis in a cohort of CD patients who underwent surgery for strictures disease Methods We retrospectively selected 48 CD patients who underwent MR-enterography and subsequent surgery within 4 months between 2017-2023. An expert pathologist evaluated the degree of fibrosis in histopathological specimens from surgical samples using the Chiorean scoring system from 0 (no fibrosis) to 2 (severe fibrosis), and the degree of inflammation using Geboes score. For each MR-enterography, typical lesions of CD were searched by an expert radiologist on T2-weighted axial images generating a region of interest (ROI) segmentation for each lesion. We extracted 100 radiomic features from each ROI of the pre-processed MRI images using Pyradiomics. Feature selection included univariate analysis (Wilcoxon-Mann-Whitney test) and Spearmann correlation to exclude highly correlated features. A logistic regression model with stepwise regression was built with the selected features and evaluated by computing the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Model’s sensitivity and specificity were calculated using the best cut-off according to the Youden index method. Results Among the 48 patients, 48% (n=23) showed severe fibrosis on the surgical specimen, whereas 52% (n=25) had either no or mild fibrosis. Two radiomic features were selected and resulted statistically significant in the fitted logistic regression model(p < 0.05): original_shape_Maximum2DDiameterColumn and original_firstorder_90Percentile. Results with their 95%CI were: AUC 0.79(0.64-0-90), sensitivity 0.69(0.50-0.75) and specificity 0.84(0.66-0.94). Bootstrap corrected (mean optimism) values of AUC and metrics were: AUC 0.77 (0.02), sensitivity 0.68(0.01), specificity 0.81(0.02). Conclusion Our study suggests that MRI-based radiomics modeling could help predict the degree of intestinal fibrosis. Future research will confirm these findings in a larger cohort

INTRODUCTION:We present the state of the art of ultrasound-based machine learning (ML) radiomics models in the context of ovarian masses and analyze their accuracy in differentiating between benign and malignant adnexal masses. MATERIAL AND METHODS:Web of Science, PubMed, and Scopus databases were searched. All studies were imported into RAYYAN QCRI software. All studies that developed and internally or externally validated ML models using only radiomics features extracted from ultrasound images were included. The overall quality of the included studies was assessed using the QUADAS-AI tool. Summary sensitivity and specificity analyses with corresponding 95% confidence intervals (CIs) were reported. RESULTS:12 studies developed ML models including only radiomics features extracted from ultrasound images, and six of them were included in the meta-analysis. The overall sensitivity and specificity for differentiating benign from malignant adnexal masses were 0.80 (95% CI 0.74-0.87) and 0.86 (95% CI 0.80-0.90), respectively, in the validation set. All studies demonstrated a high risk of bias in subject selection (e.g., lack of details on image sources or scanner models; absence of image preprocessing), and the majority also showed a high risk in the index test (e.g., models were not validated on external datasets) domain. In contrast, the risk of bias was generally low for the reference standard (i.e., most studies used a reference that accurately identified the target condition) and the testing workflow (i.e., the time interval between the index test and reference standard was appropriate) domains. CONCLUSIONS:The good performance of ultrasound-based radiomics models in the validation set supports that radiomics is worth exploring to improve the diagnosis of adnexal masses. So far, the studies have a high risk of bias due to the small sample size, single-setting design, and no external validation included.

Management of autosomal dominant polycystic kidney disease (ADPKD) might take advantage of the use of new tools to predict risk of progression towards end-stage kidney disease (ESKD). The aim of this study is to explore the potential of radiomic features obtained from computed tomography (CT) scans for the prediction of kidney function decline over time of ADPKD patients. We retrospectively selected a cohort of 58 ADPKD patients who routinely underwent CT scan for total kidney volume (TKV) assessment from February 2020 to March 2021. An expert radiologist generated a region-of-interest segmentation for cystic kidneys from which we extracted 217 radiomic features. In a subgroup of 51 patients with at least three serum creatinine measurements, on the basis of estimated glomerular filtration rate we identified 26 rapid progressors to ESKD (>3 mL/min/1.73 m2/year), and we developed a radiomic model to discriminate rapid from non-rapid progressors. Area under the curve (AUC) of the receiver operating characteristic (ROC) and sensitivity were employed to evaluate models’ performance. The most statistically significant radiomic feature (F_cm.corr) (P-value = .04) associated with rapid progression showed an AUC (95% confidence interval) of 0.78 (0.65–0.90) and a sensitivity of 0.92 (0.78–0.98). On the contrary, the logistic regression model based on the height-adjusted TKV (ht-TKV) presented a lower AUC (95% confidence interval) of 0.65 (0.49–0.80), with a sensitivity 0.62 (0.42–0.78). We developed a model based on the radiomic feature F_cm.corr that was able to discriminate rapid progressors. Further validation studies on larger and external cohort are warranted to corroborate our findings and to confirm the role of radiomics in ADPKD management.